Sie sind auf Seite 1von 2

1

Objective
The objective of this challenge is to design a feedback
controller to position the output of a linear flexible drive.
The experiment apparatus consists of one IP02 cart
followed by two linear flexible joint carts. The three carts
are coupled via linear springs and each cart is
instrumented with an encoder to measure its position
relative to the track.

Figure 2 IP02 system with two Linear flexible joints with


carts.

Substituting system parameters we obtain the following linear


model about the quiescent point (Matlab script lfjc_2l.m).

with the states defined as:

3
Figure 1 Spring
coupling between carts.
2

Control system design

The open loop system can be used to design a state feedback


controller of the form:

Figure 3 Final cart

Open loop model

The controller is designed using LQR and the resulting


The system model is derived using Maple. The file
feedback gain is:
lfjc_2l.map can be read into Maple to execute the
derivation. (Enter read `lfjc_2l.map `; in Maple). It
generates an output file lfjc_2l_m.map which is then used
in MATLAB.
4
Setting up
The model is derived using the Lagrangian formulation. A
Wire up the system as described in the table below.
simple representation is shown in Figure 4. This is used to
derive the kinetic and potential energies in the system.
UPM input
MultiQ D/A #0
The dynamic equations are derived and then a linear
model is derived by linearizing about the quiescent point.
IP02 Motor
To UPM1503 output, Gain = 3 Cable

X3
X2
X1
F

M1

M2

Figure 4 Simplified model of 3 cart system.

M3

IP02 cart
encoder

To Encoder # 0 on MultiQ

1st LFJC
encoder

To Encoder # 1 on MultiQ

2nd LFJC
encoder

To Encoder # 2 on MultiQ

Implementation and results


80

The controller is implemented using Simulink and


WinCon. The controller is drawn using Simulink - as can
be seen in the file q_lfjc_2l.mdl. The realtime code is
generated and then run in realtime using WinCon. The
sampling period chosen is 1000 Hz. The controller applies
step inputs to the system and alternates between full state
feedback control and feedback only from the SRV02. The
step responses clearly show the improvement in response
when full state feedback is used.

70

60

50

40

30

20

A sine sweep is applied to the command to obtain the


frequency response at the output with and without full
state feedback. The results are shown in Figure 5. The
resonance is clearly eliminated when full state feedback is
used. An FFT on the data also reveals removal of the
resonant peak at about 6 Hz.
The quicktime video v_lfjc_2l can also be seen using a
quicktime player.

10

10

12

14

16

18

Figure 6 FFT of output cart position response to a sine sweep


input. Green trace is without full state feedback while Blue
trace is with full state feedback.

0.15
qlfjc 2l\Amplitude

qlfjc 2l\X3

System Parameters

0.1

Symbol

Value

Units

Km

0.00767

Nm/Amp

Km (Torque
constant)

0.00767

V/(rad/sec)

Rm (Motor
resistance)

2.6

Kg (gear ratio)

3.7

NA

rp (Pinion radius)

.0063

Gear radius for


encoder

.0148

Encoder
resolution

4000

counts/turn

Calibration

2.2703e-005

m /count

Jm(Motor inertia)

3.67 e-7

Kg m2

M_1

0.7

Kg

M_2 and M_3

0.4

Kg

K_s (stiffness)

363

N/m

0.05

-0.05

-0.1

10

Time

Figure 5 Step response of final cart position. First two


responses are without full state feedback while the
subsequent responses are with full state feedback. Note
how the response is dramatically improved with full state
feedback.

20